Topdu/OpenOCR

OpenOCR: An Open-Source Toolkit for General-OCR Research and Applications, integrates a unified training and evaluation benchmark, commercial-grade OCR and Document Parsing systems, and faithful reproductions of the core implementations from a wide range of academic papers.

What it solves

OpenOCR is a comprehensive toolkit designed to handle "General-OCR" tasks. It addresses the difficulty of accurately parsing complex documents that contain a mix of plain text, mathematical formulas, and tables, while bridging the gap between academic research and industrial deployment.

How it works

The toolkit provides several specialized systems and models:

  • OpenDoc-0.1B: A lightweight document parsing system that uses a two-stage pipeline. It first performs layout analysis via PP-DocLayoutV2 and then uses the UniRec-0.1B model to recognize text, formulas, and tables.
  • UniRec-0.1B: A unified recognition model with 0.1B parameters trained from scratch on the UniRec40M dataset to recognize Chinese and English text and formulas.
  • OpenOCR System: A practical OCR system based on SVTRv2 that supports both server and mobile models, offering high accuracy for Chinese and English text detection and recognition.
  • SVTRv2: A scene text recognition benchmark and model that utilizes CTC to outperform encoder-decoder models in accuracy and speed, particularly when trained on large-scale real datasets like Union14M.

Who it’s for

It is intended for researchers focusing on OCR and document understanding, as well as developers and industry partners looking to deploy high-efficiency, commercial-grade OCR and document parsing systems.

Highlights

  • Ultra-lightweight: Includes models like OpenDoc and UniRec with only 0.1B parameters.
  • Unified Recognition: Capable of recognizing text, formulas, and tables within a single framework.
  • High Performance: OpenDoc-0.1B achieves 90.57% on OmniDocBench (v1.5), and the OpenOCR system outperforms PP-OCRv4 by 4.5% in accuracy.
  • Broad Compatibility: Supports ONNX model export for easier integration into various environments.

Related

  • Dispatch
    Mistral OCR Release Notes

    Mistral AI has launched Mistral OCR, a high-accuracy Optical Character Recognition API designed to extract interleaved text, images, tables, and equations from complex documents for use in RAG systems.

  • Project
    aiptimizer/TurboOCR

    A high-speed GPU document parser that converts images and PDFs into structured Markdown, including tables and formulas, using C++, CUDA, and TensorRT.

  • Project
    PaddlePaddle/PaddleOCR

    PaddleOCR is an open‑source OCR and document‑AI toolkit that converts images/PDFs into structured, LLM‑ready JSON or Markdown. It offers multilingual scene‑text models (PP‑OCRv6), a lightweight vision‑language parser (PaddleOCR‑VL‑1.6) for tables, formulas, and ancient scripts, and a structure‑aware pipeline (PP‑StructureV3). Models run on CPU, GPU, XPU, NPU or via ONNX/TensorRT, and integrate with RAG/agent platforms like Dify and RAGFlow.

  • Project
    scribeocr/scribeocr

    A browser-based web application for recognizing text from images, proofreading OCR data with high precision, and creating fully digitized, ebook-style documents.

  • Project
    Yuliang-Liu/MonkeyOCR

    MonkeyOCR is a document parsing model that uses a Structure-Recognition-Relation triplet paradigm to convert PDFs and images into structured Markdown, supporting English and Chinese.